Escolar Documentos
Profissional Documentos
Cultura Documentos
W. Jason Weiss1
1
Jack and Kay Hockema Professor, School of Civil Engineering, School of Materials
Engineering, Purdue University, 550 Stadium Mall Dr., West Lafayette, IN 47906.
E-mail: wjweiss@purdue.edu
Abstract
Fagerlund (1977) pioneered the use of an approach for modeling freeze-thaw damage
that considered that concrete would not be able to withstand saturation levels above a
certain value. This value has been measured to be 86% for some concrete (Li et al.
2012) and will be assumed to be 85% for this paper. A two stage absorption approach
was used to describe the rate at which concrete reached a critical level of saturation.
Bentz et al. (2011) simplified the approach assuming a single sorption rate by
assuming a weather event (e.g., rain) of longer than a particular time (6 hrs) was
required to reach the second level of saturation. Bentz et al. (2001) used this
approach with weather data to predict freeze-thaw damage in concrete pavements.
Lucero et al. (in press) used neutron radiography to illustrate that the initial sorption
was related to the filling of gel and capillary pores while the secondary rate of
sorption was related to the filling of larger pores like air voids (Todak et al. 2015).
This paper examines the role of the mixture proportions on saturation response. It is
assumed that the transition between the initial absorption and secondary absorption
(commonly referred to as the nick point) can be described by a value of saturation
that fills in all the pores in a concrete with the exception of the air voids. Powers
model (Powers and Brownyard 1942) can be used to determine the degree of
saturation at the nick point (Figure 1a) where the degree of saturation is shown as a
function of the entrained air volume. Concrete with more air has a lower degree of
saturation (i.e., is further away from reaching critical saturation). It has been shown
that when water is in contact with a small sample of concrete (50 mm thick sample)
the nick point will generally occur during the first 24 hours for typical paving
concrete (Lucero et al. in press). This value can be used with the secondary rate of
sorption to predict the time required to reach the critical degree of saturation as
shown in equation 1:
where SCR is the critical degree of saturation, SNICK is the degree of saturation
between the initial and secondary absorption which occurs when all the pores are
filled except the air entrainment, ΔS is the secondary rate of absorption and t is time.
While SNICK is dependent on the entrained air volume, the secondary rate of
absorption is strongly related to the quality of the matrix. Figure 1b illustrates the
impact of the sorption rate on the time to achieve critical saturation for a concrete
made with an air content of 5.5% and a water to cement ratios (w/c) of 0.39 to 0.45
(with SNICK varying from 66% to 70% respectively. The range of secondary sorption
was measured by Castro et al. (2011) with 0.033 being representative of a mixture
with a w/c of 0.39 and 0.053 being representative of a mixture with a w/c of 0.45.
Figure 2a provides an illustration of the influence of air content on the time it takes to
saturate the concrete. It can be seen that the volume of air has a substantial influence
on the time to reach critical saturation. Figure 2b illustrates the influence of w/c
where the impact is due to the change in the rate of water absorption.
1 40
Degree of Saturation (~)
0.8
30
0.6
20
0.4
10
0.2
0 0
0 2 4 6 8 10 0.02 0.03 0.04 0.05 0.06
Volume of Air (%) Secondary Sorption Rate (%DOS/year0.5)
(a) (b)
1.0
Degree of Saturation (~)
0.9
0.8
0.7
0.6 4% Air
6% Air
8% Air
0.5
0 10 20 30 40
Time (Years)
(a) (b)
Figure 2. a) The Illustration of the Air Content on the Degree of Saturation and
b) the Influence of Water to Cement Ratio on the Degree of Saturation
While the previous section of the paper discussed the time to reach the critical degree
of saturation for a concrete that was undamaged, this portion of the paper will begin
to discuss the role that cracking or damage may have in accelerating the rate of water
absorption. Previous research by Yang et al. (2006, 2007) has indicated that there are
two very distinctive types of crack morphology that can occur in a concrete
pavement. Crack morphology can be divided into two main types of cracking: 1)
localized discrete cracks which are consistent with mechanical loading and 2)
distributed crack networks which are consistent of those occurring due to freeze thaw
damage or alkali silica reactivity. The localized cracking can result in localized water
ingress creating regions with higher degrees of saturation. The distributed crack
networks are influenced by the crack distribution, orientation, and morphology. The
cracked networks influence the overall rate of fluid absorption. Yang et al. (2006)
showed that a cracked network that was consistent with connected cracking and a
decrease in sample stiffness of 30% resulted in a sorption that was 1.9 times higher
than that of the undamaged concrete. This is more complicated when the crack
network is not completely connected. Figure 3a illustrates the influence of damage
on the projected time to saturation. Figure 3b illustrates the projected time for the
concrete to reach critical saturation. Time to Critical Saturation (Years)
1.0 60
Undamaged
Concrete
Degree of Saturation (~)
0.7
20
0.6
Damaged Concrete
Undamaged Concrete
0.5 0
0 10 20 30 40 0 2 4 6 8 10
Time (Years) Volume of Air (%)
(a) (b)
The fluid that collects in the pavement joints is not pure water but contains deicing
salts. While the use of deicing salts is an effective method for melting the ice and
increasing the safety to the travelling public, it may also be partially responsible for
joint deterioration issues that develop at the joints in the pavement. The reactions
that occur depend on the salt that is used. It appears that NaCl results in an expansive
and damaging sulfo-aluminate phase (Farnam et al. 2014a) while CaCl2 and MgCl2
result in the formation of a calcium oxychloride (Farnam et al. 2014b, Farnam et al.
2015, Farnam et al. submitted, Sutter et al. 2008, Villani et al. 2015).
The durability of concrete pavement joints in North America have shown signs of
deterioration. This deterioration has been previously proposed by the author (2005)
to be related to fluid collecting in the joint that can lead to hydraulic pressure during
freezing and/or a chemical reaction between the deicing salt and matrix. The fluid
collects when the joint sealant fails and the joint does not crack or the crack does not
open to a sufficient width to allow it to drain. Many agencies are searching for ways
to mitigate this damage. The increase of the air content has been proposed which
may help to reduce the initial degree of saturation and increase the time for the
concrete to reach a critical saturation. This has been shown to be an effective way to
reduce the degree of saturation when the pores in the matrix saturate and this is
beneficial. It should be noted however that the volume of the air void system (and
saturation) may be compromised when the deicing salt enters the pavement reacting
with the calcium aluminate phase resulting in Friedels salt or Kuzel’s salt. It has also
been proposed that decreasing the water – to cement ratio of the pavement can
increase life by reducing the rate of fluid absorption. While both of these approaches
are valid their benefits can not completely solve the problem. It has been shown that
damage can be caused by saw-cutting (Raoufi et al. 2008) in a heart shaped region at
the base of the saw-cut. It is proposed that this damaged region could lead to an
increase in the rate of saturation and damage. By improving the timing of saw-
cutting using maturity to minimize damage at the base of the saw cut has great
potential. Coates et al. (2008) and Golias et al (2012) have suggested that the use of
a concrete sealant can reduce the rate of fluid ingress and thereby reducing the rate of
fluid ingress and potential for the concrete to reach critical saturation or the potential
chemical reaction between the salt and concrete resulting in an increase in the
durability of the concrete pavement joint. The use of supplementary cementitious
materials (SCM, fly ash, silica fume, slag) is beneficial. While use of SCM reduces
the transport properties (similar to reducing the water to cement ratio) they can also
reduce the potential for reaction between calcium chloride and calcium hydroxide.
CONCLUSION
Concrete is susceptible to freeze and thaw damage when it reaches a critical degree of
saturation. Currently, most models that describe the rate of water absorption in
concrete consider the concrete as uncracked. In service however, concrete can
develop cracks. These cracks can dramatically influence the transport of liquids and
ions. However not all cracks are created equal. This paper will discuss the rate at
which water is absorbed in concrete and the rate at which concrete reaches this
critical degree of saturation. Specifically the first portion of this paper will begin by
discussing water absorption in undamaged concrete. In the second part of the paper
the influence of cracking was discussed on the rate of water absorption. Specifically,
crack morphology will be divided into two main sources of cracking: 1) localized
discrete cracks and 2) distributed crack networks. This paper used data that shows
how distributed microcracking, similar to the cracking that is expected during freeze
thaw or alkali silica reaction, can alter the rate of absorption. Finally the potential
damage that occurs due to a reaction between the deicing salt (calcium chloride) and
the calcium hydroxide in the cementitious matrix.
ACKNOWLEDGEMENTS
This work was supported in part by the Joint Transportation Research Program
administered by the Indiana Department of Transportation (SPR 3864) and the
Portland Cement Association/National Ready Mix Concrete Association through the
CSH HUB initiative. The contents of this paper reflect the views of the authors, who
are responsible for the facts and the accuracy of the data presented herein, and do not
necessarily reflect the official views or policies of the Federal Highway
Administration, the Indiana Department of Transportation, Portland Cement
Association or National Ready Mix Concrete Association, nor do the contents
constitute a standard, specification, or regulation.
REFERENCES
Bentz, D. P., Ehlen, M., Ferraris, C., Garboczi, E., (2001) “Sorptivity-Based Service
Life Predictions for Concrete Pavements,” International Conference on
Concrete Pavements, September 2001
Castro, J., Bentz, D., & Weiss, J. (2011). Effect of sample conditioning on the water
absorption of concrete. Cement and Concrete Composites, 33(8), 805–813.
doi:10.1016/j.cemconcomp.2011.05.007
Coates, K., Mohtar, S., Tao, B., and Weiss, W. J., (2009) “Can Soy Methyl Esters
Reduce Fluid Transport and Improve the Durability of Concrete?”
Transportation Research Board, Volume 2113, pp. 22-30
Fagerlund, G. (1977). The international cooperative test of the critical degree of
saturation method of assessing the freeze / thaw. Materials and Structures,
10(4), 231–253.
Farnam, Y., Bentz, D. P., Hampton, A., & Weiss, W. J. (2014). Acoustic Emission and
Low Temperature Calorimetry Study of Freeze and Thaw Behavior in
Cementitious Materials Exposed to Sodium Chloride Salt. Journal of the
Transportation Research Record, 2441, 81–90.
Farnam. Y., Bentz, D. P., Sakulich, A., Flynn, D., and Weiss, W. J., (2014) “Using A
Low Temperature Guarded Comparative Longitudinal Calorimeter and
Acoustic Emission to Measure Freeze and Thaw Damage in Mortars
Containing Deicing Salt” Journal of Advances in Civil Engineering Materials,
ASTM, Vol. 3, No. 1, 2014, pp. 316–337, doi:10.1520/ACEM20130095
Farnam, Y., Wiese, A., Bentz, D., Davis, J. and Weiss, W. J., (2015), Damage
Development in Cementitious Materials Exposed to Magnesium Chloride
Deicing Salt, Accepted for Publication in the Journal of Construction and
Building Materials, Elsevier (In Press).
Farnam, Y., Dick, S., Wiese, A., Davis, J., Bentz, D. P. and Weiss, W. J., (Submitted)
“Influence of Calcium Chloride Deicing Salt on Phase Changes and Damage
Development in Cementitous Materials,” Cement & Concrete Composites
Abstract
INTRODUCTION
Limit states
The Ultimate Limit State (ULS) and the Serviceability Limit State (SLS) are
generally evaluated while designing/assessing a concrete structure. The general
condition for the probability of failure Pf reads:
ܲ ൌ ܲሺܵ ܴሻ ൏ ܲௗ (1)
where S is the action effect, R is the resistance (barrier) and Pd is the design
(acceptable, target) probability value. The index of reliability is alternatively utilized
instead of Pf in practice.
Generally, both S and R (and hence Pf) are time dependent and random.
Durability is related to the design working life tD (or service life) (Teplý &
VoĜechovská 2012). Two types of Pf assessment can be distinguished that consider the
randomness of concrete degradation due to biogenic sulfuric acid corrosion:
(i) Focusing on the pipe wall only, in which case the following holds for the
action effect (for a degraded concrete layer)
S=ct (2)
where c stands for the rate of corrosion; R in condition (1) represents in this case for a
value limiting the thickness of the concrete pipe wall in the sense of ULS (bearing
capacity) or SLS (deflection limit or crack opening). See Section “Parametric study”
for an example.
(ii) In a more complex case the action effect S is the bearing capacity of the
(buried) pipe (ULS or SLS) with a degraded wall, under given conditions and load;
the barrier R is a relevant limiting value.
Software tool
The FReET-D software package, a feasible and user-friendly combination of
analytical models and simulation techniques, has been used for the purposes of the
present paper. Models for carbonation, chloride ingress, reinforcement corrosion,
acid attack and frost attack are available, as well as the model for sulfate attack on
concrete in sewage collection systems shown in previous section. Altogether, 32
models are implemented as pre-defined dynamic-link library functions selected from
the literature. Fully probabilistic safety formats are employed, serving also for the
provision of quantitative information concerning a structure’s safety level. The
uncertainties associated with parameters involved in deterioration processes are
modelled by random variables. Statistical, sensitivity and reliability analyses are
provided. Several features are offered, including parametric studies and Bayesian
updating. Some of the models selected from the literature and originally developed
as deterministic models have been converted into a probabilistic form for the
purposes of this software. For more details see www.freet.cz (Novák et al 2014).
Parametric study
An ad-hoc example of the degradation of concrete sewer pipes induced by
sulfuric acid attack is described here in order to show the potential of the used model
and software. The focus is placed on the influence of two kinds of parameters വ the
wetted perimeter of the pipe wall and the type of aggregate in the concrete. Utilizing
model (4), the stochastic analysis of biogenic sulfate corrosion rate c (mm/year) is
accomplished using the input data set listed in Table 1. The wetted perimeter of the
pipe wall is varied (reflecting the filling of the pipe by 10 % വ 80 %) for two types of
concrete, which is made either from granitic aggregates (according to (ASCE 2007)
the acid-consumption capability value ranges from 0.17 up to 0.24; A = 0.20 is
chosen in present example) or calcareous aggregates (A = 0.9 up to 1.1; A = 1.0 is
used).
The resulting corrosion rates are depicted in Figure 1 (mean ± standard
deviation; the best fitted being the Gumbel max and Lognormal PDFs) demonstrating
the use of the stochastic approach.
The aggregate type appeared to influence concrete degradation to the largest
extent (Belie et al. 2004) – this effect is also clearly evident from the presented
example. Note that in reality the pipe is usually 20 % വ 40 % full, which leads in our
example to a corrosion rate at the crown of about 2 mm/year (granitic aggregate) or
0.4 mm/year (calcareous aggregate). This appears rather realistic with regard to
results mentioned in other sources; e.g. in (ASCE 2007) examples with a rate of 0.7
up to 4 mm/year for calcareous aggregates are given.
An example of reliability and assessment of type (i), Section 3.1 can be
presented together with the service life prediction as follows. Considering a limiting
value for the degraded pipe thickness of 40 mm (i.e. the barrier in Eq. (1), regarding
an SLS case with the design reliability index value ȕ = 1.5), if the reliability analysis
is carried out for granitic aggregate, the corresponding service life is about 18 years.
In the case of calcareous aggregate it would be more than 50 years.
Ring failure
The stochastic structural analysis of buried concrete sewer pipes with the
consideration of biogenic sulfuric acid corrosion is an involved, time dependant task.
Therefore, as a first step, a stochastic study for the strength analysis of the impact of
corrosion on concrete pipes has been performed, based on experimental results
presented in (Karihaloo 1995 വ Chapter 9, with reference to Gustafsson 1985).
The ring failure (or crushing strength fcr) of a non-degraded thin-walled pipe
corresponding to a single force on the top of the pipe (specifically: a linear load, Fig.
2) is studied utilizing concrete fracture theory, namely the FCM (Fictitious Crack
Model), which exhibits a good match to the test summary (2366 samples altogether!).
Note that tests on pipes with a wall thickness (t)/inner diameter (Di) ratio of 1/7 were
considered. Simple linear elasticity proved not to be relevant in this case. In this
paper the crack band model implemented in FEM was used (Atena software
ýervenka et al. 2007) to simulate the variation in the crushing strength fcr (Eq. 5,
Karihaloo 1995) of concrete pipes with different wall thickness. The ultimate
experimental top load Pu in Eq. 5 is replaced in this case by the ultimate value from
the nonlinear model.
Figure 1. Rate of corrosion vs. filling of pipe for two types of aggregate.
6 (Pu / S ) (1 + Di / D0 )
f cr = , (5)
π D0 (1 − Di / D0 )2
where Di (D0) is the inner (outer) diameter of the pipe and Pu/S is the intensity
of the ultimate linear load. Note that t/Di = 1/7. The characteristic length is defined as
EG f
lch = , (6)
ft2
The final model was consequently used to study the influence of the rate of
concrete deterioration on the variation in the crushing strength fcr of concrete pipes
with wall thickness – the results of the study can be seen in Fig. 2. The considered
reference rate of concrete deterioration was 0.5 mm/year at the crown of the pipe (see
scheme in Fig. 2). The study was carried out for 30, 60 and 90 years.
Figure 2. (a) Variation in the crushing strength fcr of concrete pipes with wall
thickness – experiments (Karihaloo 1995; black symbols), simulations (white
symbols) and the value of fcr = ft (linear elasticity prediction; horizontal line), (b)
Simulated influence of the rate of concrete deterioration on the variation in the
crushing strength of concrete pipes with wall thickness (from above 0, 30, 60
and 90 years of degradation).
Moreover, the study was also realised at the stochastic level using the approach
published in (e.g. Novák et al. 2007, Strauss et al. 2008). The rate of concrete
deterioration as well as the following dominant concrete parameters of the material
model were considered to be random variables: Modulus of elasticity, compressive
and tensile strength, and fracture energy. All of the input basic random variables
involved, and the particular set of their statistical parameters (mean value, coefficient
of variation (COV), and probability distribution function (PDF)) are summarized in
Tab. 2. Statistical correlation among the input basic random variables was also
considered (employing simulated annealing – Novák et al. 2007); see Tab. 3.
Correlation coefficients were set up based on previous experiments (e.g. Novák et al.
2007, ěoutil et al. 2014). The Latin Hypercube Sampling method (Novák et al. 2014)
was used and 16 simulations of crushing strength fcr were performed for selected
concrete pipe profiles and degradation times (0, 30, 60 and 90 years). The obtained
results – values of crushing strength fcr and their statistical parameters – are shown in
Fig. 3 and Tab. 3, and can be used as inputs in the subsequent reliability and
durability analysis of concrete pipes.
The results summarized in Tab. 3 show an interesting trend – the value of fcr for the
pipe with DN 400 (wall thickness 60 mm) after 30 years of degradation is nearly the
same as for the non-degraded pipe DN 600, etc. – see the bold numbers in the
diagonals.
Table 2. Basic random variables of concrete deterioration and dominant
material parameters.
Variable [Unit] Mean value COV PDF
Modulus of elasticity E [GPa] 37 0.1 Log-normal
Compressive strength fc [MPa] 69.5 0.1 Log-normal
Tensile strength ft [MPa] 4.52 0.12 Log-normal
Fracture energy Gf [J/m2] 210 0.2 Normal
Deterioration rate of concrete [mm/years] 0.5 0.33 Log-normal
Figure 3. Crushing strengths fcr for selected wall thicknesses (60, 100 and 140
mm) and deterioration times – 0 (a), 30 (b), 60 (c) and 90 (d) years - obtained
using a stochastic model.
Creep effect
The behaviour of the buried sewer pipe is time-dependant due to the MIC
degradation effect. Evidently the creep of concrete is an additional time-variant effect
which can play a considerable role. It can be incorporated into the above described
approach in a simplified way: at each time step a creep strain value (associated with
actual compressive stress) is added to the strain value due to mechanical loading in
each element. The statistical parameters of the creep strain can be gained using a
special FreET module which employs e.g. the B3 model.
Note that the approach described briefly in the last two sub-sections is part of
work which is currently ongoing; some results will be reported at the conference.
CONCLUSION
Acknowledgement
This work has been supported by projects No. 14-10930S and 13-22899P
awarded by the Czech Science Foundation and by project CZ.1.07/2.3.00/30.0005 –
“Support for the creation of excellent interdisciplinary research teams at Brno
University of Technology”.
REFERENCES
ASCE (American Society of Civil Engineers). (2007) Manuals and Reports of
Engineering Practice, No. 60, 2007, Gravity Sewers, NY, USA.
ýervenka, V., Jendele, L. and ýervenka, J. (2007). “ATENA program documentation
– Part 1: theory.” Prague, Czech Republic: Cervenka Consulting.
De Belie De, N. et al. (2004). “Experimental research and prediction of the effect of
chemical and biogenic sulphuric acid on different types of commercially
produced concrete sewer pipes.” Cem. Concr. Res., 34, 2223–2236.
Fernardes, I. et al. (2012). “Identification of acid attack on concrete of a sewage
system.” Mat. Struct., 45(3), 337–350.
fib Model Code 2010 (2012). fib Bulletins No. 65 and 66. International Federation for
Structural Concrete, Lausanne, Switzerland.
ISO 13823 (2008). General Principles on the Design of Structures for Durability.
1
Department of Civil and Environmental Engineering, University of Pittsburgh, PA
15261. E-mail: qiy15@pitt.edu
2
Department of Civil and Environmental Engineering, University of Pittsburgh, PA
15261. E-mail: tet16@pitt.edu
Abstract
One of the challenges in bridge engineering is to predict the long-term deflection and
damage accumulation in large-span prestressed concrete girders during their lifespan.
For prestressed concrete bridges, a salient by-product of time-dependent deformation
resulting primarily from concrete creep is unexpected stress redistribution, which
triggers damage and cracking in the critical structural components. To deepen the
understanding of the coupled effects of deflection and damage on bridge
performance, a case study is carried out in this investigation, in which a large-span
prestressed concrete bridge is modeled in ABAQUS based on its blueprints available.
To capture the concrete behavior during service, a constitutive law formulated based
on a damage model is employed to approximate the nonlinear softening after concrete
cracking. This formulation is further integrated with a rheological model resting on
Kelvin units for concrete creep. Using a rate-type algorithm, the long-term behavior
of the bridge is probed.
INTRODUCTION
With the increasing demands in construction sustainability, a longer lifespan, i.e.,
over 100 years, is now generally expected for critical prestressed concrete girders. A
serious challenge to this goal is the low predictive capacity of the current design
models for bridge long-term performance, a complex outcome of time-dependent
interactions between concrete creep, shrinkage, steel relaxation, and deterioration
processes (Strauss et al. 2013, Wendner et al. 2010), coupled with other physical and
mechanical influences.
ε iji ε ijt
where ε iji is the instantaneous strain after loading, and ε ijt is its time-dependent
counterpart consisting of inelastic strain induced by concrete static creep ε ''ij , cyclic
creep ε ijcc and shrinkage ε ijsh . In this study, autogeneous shrinkage is neglected and
only the drying shrinkage resulting from moisture loss is taken into account.
RATE-TYPE FORMULATION
Within the service stress level, concrete static creep is assumed to follow the
constitutive law of ageing linear viscoelasticity (Jirásek and Bažant 2002, Bažant et
al. 2012 a, b). Due to concrete ageing, the compliance function J(t, t’), serving as
kernel in Volterra integral form, is not of a convolution type. Therefore, it may be
approximated by a rheological model in the rate-type formulation so as to capture the
strain growth under a general stress history σ (t).
If Kelvin units are employed, the compliance function can be approximated as
(Hardy and Riesz 2013, Schapery 1962, Jirásek and Bažant 2002):
1 M
1 t − t′
J (t , t ′) ≈ +¦ [1 − exp(− )]
E (t ) i =1 Di (t ′) τi
where t is current time and t’ is the loading age. To avoid thermodynamically
inadmissible Di obtained through data fitting, continuous spectrum method is adopted
to uniquely and efficiently identify the compliance spectrum Ai. Utilizing Laplace
transformation inversion supplemented by Widder’s approximate inversion formula
(Widder 1971), the analytical solution of Ai for a given compliance function can be
attained; see Yu et al. 2012.
Damage in concrete can be approximated by a number of forms based on
different formulation. Here a damage model resting on an isotropic damage variable
Φ will be employed (Kachonov, 1958). Using the strain equivalence hypothesis and
generalized Hook’s law, the effective (undamaged) stress tensor σij can be expressed
as:
σ ij = Eijkl (t )ε kle = Eijkl (t )ε kle
where E ijkl (t) is the fourth-order effective (undamaged) isotropic elasticity tensor,
expressed as a function of concrete age t. After concrete cracking, concrete stress-
strain relation can be described by a simplified softening law as:
σ ij = (1 − Φ)σ ij = (1 − Φ) Eijkl (t )ε ije
here the evolution of damage variable Φ follows an isotropic damage rule hinging on
the critical principle strain of concrete. In the formulation, this damage variable will
be assigned to all Kelvin units based on the assumption that same degree of damage
happens in each Kelvin unit.
For the cyclic creep, a power law-type equation similar to Paris law will be
used. The development of cyclic creep after N cycles is approximated as (Bažant and
Hubler, 2014):
Δσ m
ε cc = C1σ ( ) N
fc′
Here f c' is the compressive strength of concrete, Δσ is the amplitude of stress
variation and σ is the average stress in concrete. Note that ε cc depends on both σ
and N linearly, which agrees with the test measurements available in literature and is
convenient for structural analysis. In this study, the exponent value m is set as 4, and
C1 is about 46×10-6.
For a rate-type formulation, the 3D quasi-elastic stress-strain incremental
relation at any time step can be written as:
Δσ = E′′Δε e
where E '' is the effective incremental modulus considering viscoelasticity and
concrete cracking. The increment of the effective elastic strain Δεe is:
Δε ije = Δε ij − Δε ijcc − Δε ijsh − Δε ij′′
(1) For the prestressed concrete bridges carrying heavy traffic flow cyclic creep of
concrete must be taken into account in bridge design. The neglect of cyclic creep
may lead to unrealistic long-term deflection prediction.
(2) Due to the time-dependent deformation, cracking will happen in concrete as a
consequence of stress redistribution. The interaction between concrete damage
and creep will augment the time-dependent deflection of prestressed bridges.
(3) To capture the coupled effect of damage and time-dependent deformation, rate-
type formulation is needed. In addition to cyclic creep and concrete cracking, it
can be further extended to incorporate other physical and chemical processes,
e.g., corrosion.
REFERENCES
ACI Committee (2008). Building Code Requirements for Structural Concrete (ACI
318-08) and Commentary. Farmington Hills: American Concrete Institute.
Bažant, Z. P., and Prasannan, S. (1989 a). “Solidification theory for concrete creep II:
verification and application.” Journal of Engineering mechanics, 115(8),
1704-1725.
Bažant, Z. P., and Prasannan, S. (1989 b). “Solidification theory for concrete creep I:
formulation.” Journal of engineering mechanics, 115(8), 1691-1703.
Bažant, Z. P., and Baweja, S. (2000). “Creep and shrinkage prediction model for
analysis and design of concrete structures: Model B3.” ACI SPECIAL
PUBLICATIONS, 194, 1-84.
Bažant, Z. P., Hubler, M. H., and Jirásek, M. (2012). “Improved estimation of long-
term relaxation function from compliance function of aging concrete.” ASCE
J. of Engrg. Mech, 139 (2), 146–152.
Bažant, Z. P., Yu, Q., & Li, G. H. (2012 a). “Excessive long-time deflections of
prestressed box girders. I: Record-span Bridge in Palau and other paradigms.”
Journal of Structural Engineering, 138(6), 676-686.
Bažant, Z. K. P., Yu, Q., & Li, G. H. (2012 b). “Excessive long-time deflections of
prestressed box girders. II: Numerical analysis and lessons learned.” Journal
of Structural Engineering, 138(6), 687-696.
Bažant, Z. P., Hubler, M. H., and Jirásek, M. (2013). “Improved Estimation of Long-
Term Relaxation Function from Compliance Function of Aging
Concrete.” Journal of Engineering Mechanics, 139(2), 146-152.
Bažant, Z. P., & Hubler, M. H. (2014). “Theory of cyclic creep of concrete based on
Paris law for fatigue growth of subcritical microcracks.” Journal of the
Mechanics and Physics of Solids, 63, 187-200.
Brooks, J. J. (1984). “Accuracy of estimating long-term strains in concrete.”
Magazine of Concrete Research, 36(128), 131-145.
Brooks, J. J. (2005). “30-year creep and shrinkage of concrete.” Magazine of concrete
research, 57(9), 545-556.
Abstract
1. INTRODUCTION
© ASCE
CONCREEP 10 239
© ASCE
CONCREEP 10 240
manifested in the time measured for days and years, which creates problems for their
joint consideration.
σ (t ) t
∂ 1
t
∂
ε (t ) = − ³ σ (τ ) dτ − ³ σ (τ ) C ( t ,τ ) dτ , (1)
E (t ) τ1 ∂τ E (τ ) τ1 ∂τ
1 d 1
εe (τ ) = σ (τ ) + σ (τ )
E (τ ) dτ E (τ )
© ASCE
CONCREEP 10 241
σ (τ 1 ) t 1
t
d 1
ε e (τ ) = + ³ σ (τ ) dτ + ³ σ (τ ) dτ .
E (τ 1 ) τ1
E (τ ) τ 1
dτ E (τ )
σ (τ 1 ) t 1
ε e (τ ) =
E (τ 1 ) τ³ E (τ )
+ dσ .
1
The error is a loss of the deformation portion, which corresponds to the rate
of change of stiffness. The same error exists in the last integral term. Under the
conditions of nonlinear creep further this loss leads to the formulation of the
principle of superposition strange that violates not only the principles of Newtonian
mechanics, but also the conditions of the affine scaling experimental creep curves.
Consider the third term in (1), and write with the help of that part of the linear creep,
which is called instantaneous.
t
∂
t
∂C (σ , t ,τ ) t
∂C (σ , t ,τ )
ε n ( t ) = ³ σ (τ ) Cn (σ , t ,τ ) dτ = ³ σ (τ ) σ (τ ) n dτ + ³ σ (τ ) n dτ .
τ1 ∂τ τ1 ∂σ τ1 ∂τ
However, the calculation of the instantaneous creep does not operate, and the
integral is usually written in the form
t
∂
ε n ( t ) = ³ σ (τ ) Cn ( t ,τ ) dτ . (2)
τ1 ∂τ
Here are the values ε n (t ) , linearly depends on the stresses, that corresponds
neither to experiment to find ε n (t ) , or the data in Fig. 1 of the nonlinear coupling
ε n and ı.
To describe Cn(t,IJ) used a variety of complex formulas that do not meet the
obvious experimental data. For example, in the approximation of the diagram σ − ε M
square parabola, we have the exact value of the deformation ε n = β 2σ 2 :
t
∂
ε n ( t ) = β 2σ 2 ( t ) = ³ σ (τ ) Cn ( t ,τ ) dτ ,
τ1 ∂τ
which should correspond to the right side, including its function C(t, IJ), that it is
impossible to perform traditional records, for example, at Alexandrovsky S.
© ASCE
CONCREEP 10 242
1 − A2 e −γτ −γ ( t −τ )
C ( t ,τ ) = ψ (τ ) −ψ ( t ) e + Δ (τ ) ª¬1 − e −γ ( t −τ ) º¼ .
1 − A2 e −γ t
Complex formulas in the description of C(t, IJ), designed to take into account
the instantaneous creep significantly increase the order of the corresponding
differential equation of creep of concrete. This complicates the solution of practical
problems of ordinary calculation of reinforced concrete structures.
Update the last term in (1), using the properties of potential forces under
creep conditions (Sanzharovsky R., 2014). Find the velocity of creep deformation-
∂C ( t ,τ ) ∂C ( t ,τ )
εc ( t ,τ ) = σ (τ ) C ( t ,τ ) + σ (τ ) + σ (τ ) ,
∂τ ∂t
t t
∂C ( t ,τ ) t
∂C ( t ,τ )
ε c ( t ) = σ (τ 1 ) C ( t ,τ 1 ) + ³ σ (τ ) C ( t ,τ ) dτ + ³ σ (τ ) dτ + ³ σ (τ ) dτ .
τ1 τ1 ∂τ τ1 ∂t
t
∂C ( t ,τ )
ε c ( t ) = ³ σ (τ ) dτ . (3)
τ1 ∂t
The last term in (1) and the value of (3) in linear creep may not differ from
each other only by using the difference of the nuclei. This fact characterizes the
corresponding superposition principle, as well as the correctness of its further
application in the nonlinear creep.
These unauthorized principles and errors have a significant impact on the
results of the calculations of reinforced concrete structures creep. They also show
that in the conventional form of (1) the equation for creep regulation and mass use in
reinforced concrete structures is unacceptable.
Previously, the authors of this paper (Sanzharovsky R., 2013) proposed a
method of accounting instantaneous nonlinearity of concrete creep. According to the
hypothesis of linear concrete creep basic equation by creep can be written as
t
∂
ε ( t ) = f 2 ª¬σ ( t ) º¼ + ³ f1 ª¬ε M (τ ) º¼ C ( t ,τ ) dτ , (4)
τ1 ∂t
© ASCE
CONCREEP 10 243
aε M + δε M2 a σ + δ 0σ 2 1
σ = f1 ( ε M ) = ; ε M = f 2 (σ ) = 0 ; a = EB ( t ) ; a0 = ;
c1 + gε M c0 + g0σ EB ( t )
Rδ ( t ) EB ( t ) ε B1 − 2 RB ( t )
δ =− ; g= ; c1 = c0 = 1 and so on.
ε 2
B1 RB ( t ) ε B1
σ = f1 ª¬ε M ( t ) º¼ = EB ( t ) ε M − A2 ( t ) ε M2 , (5)
1
εM = σ + B2 ( t ) σ 2 = f 2 ª¬σ ( t ) º¼ ,
EB ( t )
where the differential equation will have a second order. Measure the creep of
the Alexander S. lead to differential equations of the fifth order. Pay attention to the
important requirement to the function C(t,IJ). The equations of second and higher
orders appear drag force proportional to the total acceleration ε(t ) . This is
incompatible with Newtonian mechanics and violates its fundamental principle of the
independence of the forces.
Saving the form action creep Harutyunyan N., Prokopovich I., Karapetyan K.,
we obtain the differential equation of the first order that is convenient for practical
tasks
a ( t ) ε M ( t ) + δ ( t ) ε M2 ( t )
ε ( t ) + γε ( t ) = εM ( t ) + γθ ( t ) + γε M ( t ) , (7)
1 + g (t ) ε M (t )
or
© ASCE
CONCREEP 10 244
Esec ª¬σ ( t ) º¼
Esec ª¬σ ( t ) º¼ ª¬ε ( t ) + γε ( t ) º¼ = σ ( t ) + γ {1 + θ ( t ) Esec ª¬σ ( t ) º¼}σ ( t ) . (8)
Etan ª¬σ ( t ) º¼
Let’s pay attention to the important results. The presence in the creep law (7)
instantaneous deformation İɦ concrete and its relation to the total strain İ opens
opportunities for almost an exact solution of a number of fundamental design
problems, such as long-term resistance to compression of concrete columns and
beams ( Sanzharovsky R., 1978). It is possible, under the conditions of nonlinear
creep and instantaneous nonlinearity of concrete, to take into account the exact stress
distribution over the cross section within the deformation model of Eurocode 2, and
thereby bridge the gap that exists between the theories of short-term and prolonged
resistance reinforced concrete structures. It is possible in the evaluation of long-term
ultimate bearing capacity of structures to apply the criterion to achieve the
normalized value of the limiting compressive deformation İɜ2 Eurocode 2. Parallel
accounted for the change over time of concrete strength RB.
3. EXAMPLE
ba c
° N = g 2 ε ª¬ gε 2 ɦ − ln (1 + gε 2 M ) º¼ +
° 2M
° bδ c ª 1 2 2 º
°+ 3 « g ε 2 M − gε 2 M + ln (1 + gε 2 M ) » +
° g ε 2M ¬ 2 ¼
° ' ' ε2 ε2
° + As Es ( c − a ) − As Es ( h0 − c ) ,
'
° c c
® (9)
° M = N ( e + f ) = bδ c ª 1 g 3ε 3 − 1 g 2ε 2 + gε − ln (1 + gε ) º +
2
g 4 ε 22M «¬ 3
2M »
° 2M
2
2M 2M
¼
°
° + bδ c ª 1 g 2ε 2 − gε + ln 1 + gε ' ' ε2
2
º
( c − a' ) +
2
° g 3 ε 2 «¬ 2 2M 2M ( 2 M ) » − As Es
¼ c
° 2M
° ' ' ε2
°̄+ As Es c ( h0 − c ) ,
2
© ASCE
CONCREEP 10 245
∂N ∂N ∂N
° ∂ε ε 2 ( t ) + ∂c c ( t ) + ∂ε ε2 M ( t ) = 0,
° 2 2M
® (10)
°N ∂f ∂ f ∂M ∂M ∂M
ε2 ( t ) + N c ( t ) = ε2 ( t ) + c ( t ) + ε2 M ( t ) = 0.
°¯ ∂ε 2 ∂c ∂ε 2 ∂c ∂ε 2 M
ε2 ( t ) = Φ 2 ª¬ε 2 ( t ) , ε 2 M ( t ) , c ( t ) º¼ ,
ε2 M ( t ) = Φ 2 M ª¬ε 2 ( t ) , ε 2 M ( t ) , c ( t ) º¼ , (11)
c ( t ) = Φ c ª¬ε 2 ( t ) , ε 2 M ( t ) , c ( t ) º¼ .
∂N ∂N ∂N
° ∂ε δε 2 + ∂c δ c + ∂ε δε 2 M = 0,
° 2 2M
° ∂f ∂f ∂M ∂M ∂M
®N δε 2 + N δ c = δε 2 + δc + δε 2 M = 0, (12)
° ∂ε 2 ∂c ∂ε 2 ∂c ∂ε 2 ɦ
°δε 2 = δε 2 M .
°
¯
© ASCE
CONCREEP 10 246
Determinant of the system (12) is equal to zero, is the condition of the critical
state. This determinant is calculated together with the system of equations (11). It
captures the moment of the ultimate state, the time loss of bearing capacity of the
structure, including the duration of the existence of the structure.
4. CONCLUSIONS
© ASCE
CONCREEP 10 247
REFERENCES
Sanzharovsky, RS 1978, Stability of construction elements during creep,
LSU. (in Russian)
Sanzharovsky, RS 2013, 'Problems of the theory of creep', Structural
Mechanics engineering structures and buildings, no. 3, pp. 28-34. (in Russian)
Sanzharovsky, RS 2014, 'Nonlinear hereditary creep theory', Building
mechanics engineering structures and buildings, no. 1, pp. 63-68. (in Russian)
.
© ASCE
CONCREEP 10 248
Abstract
Most of large civil engineering concrete structures have been instrumented for
decades with embedded sensors. To prevent the eventual loss of data, complementary
instrumentation of external surface has recently been deployed. This new
instrumentation can take different forms but in all cases, to avoid damaging the
structure, it will be only superficially anchored. Near the outer surfaces, thermo-hydro-
mechanical concrete behaviour is more sensitive to varying environmental conditions
than in the centre of the structures. Therefore, the strain measured near the outer surfaces
is not identical to the strain measured by embedded sensors. Consequently the methods
of classical physical-statistical analysis must be reviewed. Using a thermo-hydro-
mechanical finite element modeling calibrated on a representative concrete and applied
on a current part of a thick structure, this work confirms a dependence of strain on the
depth. First results show that the depth impact affects both kinetic and amplitude strain.
Keywords:
Thermo-hydro-mechanical behavior; Concrete; Massive structure; Variable
environnemental conditions; Surface monitoring.
INTRODUCTION
For nearly 80 years, most of large civil engineering concrete structures are
instrumented with vibratory strain gauge. On some structures such as confinement
vessels of nuclear power plants (highly prestressed), operators have to cope with the
© ASCE
CONCREEP 10 249
loss of signal from some sensors. This can be explained by several reasons including a
default in the electrical system that excite the wire or by the excess of frequency limit
measurement due for instance to important creep strains (Simon and Courtois, 2011;
Simon et al. 2013). Most of these sensors have been embedded in the concrete during
construction and are thus irreplaceable. To overcome this loss of information, one of
the solutions is to replace these defective embedded sensors by surface sensors
(Figure 1).
Figure
1:
Embedd
ed strain
sensor
(a) and
surface
strain
sensor
(b).
Nevertheless, unlike central strain (far from outer surface), skin strain (near outer
surfaces) is strongly affected by the environmental condition variations. Temperature
and humidity cycles (daily and seasonal) generate thermo-hydric variations of volume.
Other phenomena such as sunshine, rain or wind, also disrupt these two main fields.
Finally, characteristic times and so influence depths of these phenomena are very
different. For these reasons, the analysis of this new data is more complex than
embedded measures for which usual statistical analysis methods exist.
The aim of the present contribution is to compare the strain evolution through the
depth in thick structure. This can be considered as a first step toward the complete
definition of a space and time transfer functions between both strain vs time signals.
For this purpose, a visco-elastic finite element model able to represent thermo-hydro-
mechanical behaviour of concrete submitted to variable thermo-hydric boundary
conditions is presented in the first part. Then, the calibration of the material constitutive
parameters on laboratory tests is detailed. Finally, the strains at different depths from a
simulation over thirty years on a current part of containment enclosure submitted to
thermo-hygrometric variations are analysed.
© ASCE
CONCREEP 10 250
NUMERICAL MODEL
Numerical strategy.
To be applicable to major civil engineering thick structures, the numerical model must
account for thermal and hydric variation effects on strains. Thermo-mechanical
(Arthanari and Yu, 1967 ; Seki and Kawasumi, 1972 ; Kommendant et al, 1976),
hydro-mechanical (Wittmann, 1970 ; Wittmann, 1973) and thermo-hydric (Caré,
2008) chaining are retained. In contrast, the effects of mechanical strains on thermal
and hydric properties (Lassabatere et al, 1997) and the modification of heat and water
diffusivity due to cracked skin are neglected. Finally, for ageing structures, hydration
reaction is supposed finished.
Complete simulations are carried out using Code_Aster by chaining three calculations:
thermal field, hydric field and mechanical strains. Constitutive models adopted are for
the most part empirical although based on physical approaches.
Thermal field.
Thermal field is assumed to be governed by a conventional linear heat equation with
convective boundary conditions (Neumann type):
{
ȡ Cth TÚ = k th ǻT
ࢥ th = hth( T í T )
(1)
Specific heat capacity ȡCth, thermal conductivity kth and heat transfer coefficient hth
(taking into account convection and radiation) are taken constant during all the
calculation.
Hydric field.
Drying process is modeled by a nonlinear diffusion equation involving a diffusion
coefficient (non-linear function of water content) usually written as:
CÚ = 䳱[ D(C , T ) 䳱T ] (2)
Several authors have proposed various relationships between drying coefficient and
water content. The expression used, proposed by Granger (1995), is composed of the
drying coefficient proposed by Mensi (1988) and the temperature dependence given
by Bazant (1972):
íQ 1 1
( í )
bC T R T T0
D(C , T ) = a e e (3)
T0
© ASCE
CONCREEP 10 251
The coefficients a,b and the activation energy Q of Arrhenius' law are supposed to be
constant. T0 represents the reference temperature.
The calculation of drying is done by means of water content, although boundary
conditions are prescribed in relative humidity. The isotherm between relative humidity
of concrete and water content must be entered. In this study, the hysteresis existing
between sorption and desorption cycles is neglected and the isotherm is supposed
linear in the humidity range between 40% and 100%. Therefore, the used isotherm
function is:
C eq [h 0í h] í C0 [ hí heq ]
C( h) = (4)
h0í heq
where h0, C0 are the relative humidity and initial moisture content and heq, Ceq are the
relative humidity and water content at equilibrium.
The convective boundary condition used involves a water exchange coefficient hdr
which is assumed constant:
ࢥ dr = hdr [ Cí C( hair )] (5)
By analogy with nonlinear thermal calculation, water field is obtained using the non-
linear thermal module of Code_Aster.
Mechanical field.
This study focuses on operating structures during service life. Consequently damage
rate is assumed to be low and a simple elastic model is used.
Under the assumption of small strains, the total strain tensor is assumed to be the
sum of five tensors: elastic strain (İEl), thermal strain (İTh), drying shrinkage strains
(İDS), basic creep strains (İBC) and drying creep strains (İDC):
İtot = İEl + İTh + İDS + İ BC + İ DC (6)
© ASCE
CONCREEP 10 252
CALIBRATION
Thermal field.
The thermal diffusion and the specific capacity cannot be measured out of these tests.
Thus, their retained values were obtained from the literature. It is the same for the
heat exchange coefficient.
Hydric field.
Before calibrating the drying model, the sorption-desoprtion isotherm of the concrete
must be determined in order to translate the boundary conditions (known in relative
© ASCE
CONCREEP 10 253
Elastic strain.
Measurements from the first campaign were used to determine the Young's modulus of
concrete studied. A linear regression of the stress-strain curves (Figure 3a) allows to
determine an average Young's modulus for these four specimens.
Since only axial strains were measured, the Poisson's ratio could not been
determined. Therefore a standard value for concrete is used.
Thermal strain.
No thermal expansion test was made. But peaks of temperature up to 25°C were
measured during the campaign. These variations have been felt on shrinkage and
creep tests. These thermal anomalies have been used to calibrate the thermal
expansion coefficient.
The analysis of the experimental results yields an increase of strain with temperature
amplitude (Figure 3b). A linear regression of these points gives direct access to the
thermal expansion coefficient ĮTh.
© ASCE
CONCREEP 10 254
Drying shrinkage.
The drying shrinkage strain is assumed to be proportional to the variation of water
content. Therefore, only one parameter must be identified (ĮDS). Autogenous
shrinkage was estimated at short-term from non-drying weight loss. Then, it was
substracted to drying shrinkage test in order to obtain the drying shrinkage. The
obtained curve is not exactly linear at the beginning (Figure 3c). Indeed, at the start
of drying, the high tensile stress located near the exchange surfaces creates skin
cracking, which relaxes the stress.
A linear regression of the central portion of the curve allows the calibration of the
drying shrinkage parameter. Numerically, the strain obtained will be greater than the
experimental measurement. However both kinetic and amplitude related to relative
humidity cycles will be respected.
Drying creep.
It was decided arbitrarily to calibrate drying creep before basic creep (since it has
only one parameter). Experimental curve of drying creep was obtained by subtracting
non-drying creep and drying shrinkage to the drying creep test. Since drying creep
was assumed proportional to the stress and the relative humidity variation, drying
creep values were divided by the applied stress. Meanwhile, the mass loss was
translated into relative humidity thanks to the sorption-desorption isotherm.
A quasi-linear increase of drying creep with relative humidity variation is highlighted
by the analysis of experimental results in Figure 3d. Again, as for drying shrinkage,
the resulting curve is not perfectly linear in its ends. For the beginning, explanations
may be the same than in the case of drying shrinkage. For the long term, the
© ASCE
CONCREEP 10 255
Basic creep.
Finally, basic creep is calibrated on the total creep test, taking into account the
previously calibrated drying creep. Experimental and numerical results are presented in
Figure 2b.
SIMULATION METHOD
© ASCE
CONCREEP 10 256
double containment vessel is studied which allows the suppression of the direct effects
of rain, wind and sunshine; only thermal and humidity variations, load and prestressing
were selected as thermo-hydro-mechanical boundary conditions.
The mesh of a current portion of containment in three dimensions (width of 2 meters,
height of 1.8 meters, thickness of 1.2 meters and internal radius of 21.9 meters),
taking into account reinforcements was used (Figure 4).
The calculation simulates 30 years of service life of the structure. The thermo-
hygrometric boundary conditions were created by duplicating periodic temperature
and humidity measurements from the studied structure.
During the first three years, only the structural loading, the thermal and the moisture
variations (identical on the inner and outer faces) are taken into account.
At 3.5 years, the vertical and horizontal preloading is added.
At 6 years, the reactor is operating. This yields a change in thermo-hygrometric
boundary conditions (different on the inner and outer sides).
At 16 and 26 years, two containment pressure tests are simulated (pressurizing the
reactor vessel for 3 days). For that, an overpressure is added on the inner side and
an depression is added on the upper face (to represent the vertical stretch of the
structure).
© ASCE
CONCREEP 10 257
RESULTS :
The orthoradial strains obtained numerically on the outer surface and near the
center of the structure are presented in Figure 5. Also for confidentiality reasons, the
strains are normalized by the absolute value of the center strain at 30 years.
Figure 5: Tangential strain on outer surface and close to center of structure.
CONCLUSION
The results of this study show that, unlike thermal variations which affect the
entire thickness of the structure, creep kinetics are not the same between the center and
the surface of the structure. Creep is smaller near the exchange surfaces, where
© ASCE
CONCREEP 10 258
Work performed thanks to the support of EDF in context of Chair PERENITI run by
the Fondation Partenariale Grenoble INP and using the monitoring data provided by
EDF-DTG. Partners responsibility of the Chair cannot in any circumstances be
blamed on the grounds of the content of the publication, which is only binding its
author.
REFERENCES
Arthanari S., Yu C.W. (1967). “Creep of conrete under uniaxial and biaxial stresses
at elevated temperatures”. Mag. Concrete Res. 13(60), 149-156.
Bazant Z.P., Najjar L.J. (1972). “Nonlinear water diffusion in nonsaturated concrete”.
Materials and Constructions, 1972, Vol. 5, 3-20.
Bazant Z.P., Chern J.C. (1985). “Concrete creep at variable humidity: constitutive
law and mechanism”. Matériaux et Constructions, Vol. 18, n°103, pp 1-20.
Caré S. (2008). “Effect of temperature on porosity and on chloride diffusion in cement
pastes”. Construction and Building Materials 22, 1560-1573.
Code_Aster, Électricité De France (EDF), Code_Aster finite element code version
11.5, available at http://www.code-aster.org.
Granger L. (1995). “Comportement différé du béton dans les enceintes de centrales
nucléaires: analyse et modélisation”. Laboratoire Central des Ponts et
Chaussées.
Kommendant G., Polivka M., Pirtz D. (1976). “Study of concrete properties for
prestressed concrete reactor vessels”. Rapport technique, Department of Civil
Engineering.
Lassabatere T., Torrenti J.-M., Granger L. (1997). “Sur le couplage entre séchage du
© ASCE
CONCREEP 10 259
© ASCE